Soil selection for use in environmental analysis

ABSTRACT:

Estimating variables, such as erosion rates, across a diverse soil and water resource base is a problem of interest in natural resource management. Here, we propose an alternative to the predominant soils (PS) approach of Stoneman, Brown, and Spivey. Our method called Gaussian quadrature (GQ), is adapted from the numerical integration literature. Two experiments compare GQ with PS. The first plugs sample input into a simulation model to approximate erosion rates, surface runoff, and crop yield for a region. The second estimates erosion in three regions using the Universal Soil Loss Equation (USLE). Results for the GQ samples are compared to results for the full population and results from random samples. GQ sampling tends to be more effective, particularly with respect to measures related to the heterogeneity of the population, such as the variance and skewness, than PS or random sampling. Judiciously used GQ sample selection permits reductions in the number of soils sampled with only a moderate loss of accuracy.